How phuc-nt/my-translator works
My Translator is a Tauri desktop app that captures system and/or microphone audio, translates speech through Soniox, OpenAI Realtime, Qwen, or offline MLX, and presents results in an overlay. The main path runs from user audio selection through capture and engine-specific translation to the transcript UI; optional narration and session history are separate branches. README documents broad provider and UI capabilities, while sampled source confirms only some wiring, so unverified subsystem links are omitted. The graph focuses on runtime behavior, not build or setup machinery.
Main entry points: app.js App controller · audio.rs Capture commands · openai_realtime.rs OpenAI realtime · lib.rs Tauri runtime
Components 26
Audio capture
- Capture commands
src-tauri/src/commands/audio.rs - Microphone capture
src-tauri/src/audio/microphone.rs - System audio capture
src-tauri/src/audio/system_audio.rs - Windows loopback
src-tauri/src/audio/wasapi.rs - Audio resampler
src-tauri/src/audio/resampler.rs
Translation engines
- Soniox translation
src/js/soniox.js - OpenAI client
src/js/openai-realtime-client.js - OpenAI realtime
src-tauri/src/commands/openai_realtime.rs - Qwen client
src/js/qwen-realtime-client.js - Qwen realtime
src-tauri/src/commands/qwen_realtime.rs - Local MLX pipeline
src-tauri/src/commands/local_pipeline.rs
Overlay and history
- App controller
src/js/app.js - Settings
src/js/settings.js - Tauri runtime
src-tauri/src/lib.rs - Transcript overlay
src/js/ui.js - Session manager
src/js/session-store.js - Session persistence
src-tauri/src/commands/session_store.rs - Local transcripts
Speech output
- Read-aloud reader
src/js/reader.js - TTS providers
src/js/elevenlabs-tts.js - Audio playback
src/js/audio-player.js
Other components
- User
- Soniox service
- OpenAI service
- Qwen service
- Local models
Connections 27
- User → App controller: configures and starts
- App controller → Settings: reads settings
src/js/app.js - App controller → Capture commands: starts capture
- Tauri runtime → Capture commands: registers commands
src-tauri/src/lib.rs - Capture commands → Microphone capture: captures microphone
- Capture commands → System audio capture: captures system audio
- System audio capture → Windows loopback: uses loopback
- App controller → Soniox translation: routes translation
src/js/app.js - Soniox translation → Soniox service: streams audio and text
- App controller → OpenAI client: routes translation
src/js/app.js - OpenAI client → OpenAI realtime: invokes commands
src/js/openai-realtime-client.js - OpenAI realtime → Audio resampler: upsamples audio
src-tauri/src/commands/openai_realtime.rs - OpenAI realtime → OpenAI service: streams audio and translation
- App controller → Qwen client: routes translation
src/js/app.js - Qwen client → Qwen realtime: invokes commands
src/js/qwen-realtime-client.js - Qwen realtime → Qwen service: streams audio and text
- App controller → Local MLX pipeline: starts local mode
- Local MLX pipeline → Local models: runs offline inference
- App controller → Transcript overlay: updates transcript
src/js/app.js - App controller → Read-aloud reader: starts narration
src/js/app.js - Read-aloud reader → TTS providers: requests speech
src/js/app.js - TTS providers → Audio playback: plays generated audio
- App controller → Session manager: records segments
src/js/app.js - Session manager → Session persistence: saves and reads sessions
src/js/session-store.js - Session persistence → Local transcripts: persists files
src-tauri/src/commands/session_store.rs - Tauri runtime → OpenAI realtime: registers commands
src-tauri/src/lib.rs - Tauri runtime → Qwen realtime: registers commands
src-tauri/src/lib.rs